European Economic Integration and EU Cohesion Policy

(1)

European Economic Integration and EU

Cohesion Policy

An Econometric Approach based on Input-Output Data

Abstract

This thesis combines data on value chain fragmentation for 28 EU countries with the cohesion investments under the EU Regional Policy for the budgetary perspective from 2007 to 2013. The outcomes of the Input-Output computations update the findings of an existing study by Los et al. (2015) and suggest that the foreign value added shares increase during the observation period. A regression analysis with the structural funds as predictor for the value added fragmentation reports a marginal impact on the cross-border production linkages, especially in the same or the subsequent year of funding. Due to unsolved causality problems, the conclusiveness of the results is limited.

Keywords: Value Added Fragmentation, Cohesion Politics, Global Value Chains

submitted by

Anton F

USSSTETTER

20.06.2017 E-Mail: anton.fussstetter@web.de ID (RUG): S3215733 ID (GAU): 21565783 Phone: +31 621 688 159 Address: Ebrachweg 2 83533 Edling Germany

1stSupervisor: Prof. Dr. Marcel.P. TIMMER

(2)

Chapter I

Introduction

The budget for investments under the Regional Policy of the European Union (EU) sur-passed the expenditure for the Common Agricultural Policy for the first time in the 2007-2013

financial perspective, making it the largest fraction of the EU budget commitments (

Euro-pean Commission, 2006). Beyond that, the planned expenses for the current programming

period 2014-2020 increased again by about one third compared to the previous period (

Euro-pean Commission,2014). Due to this substantial devotion of financial resources, the Regional Policy is of particular importance for the implementation of EU objectives. Therefore, the evaluation of structural funding as the most important investment policy of the EU has a pre-eminent role in assessing the functioning of the union.

The EU is undergoing constant change and the recent enlargements considerably widened the wealth gap between the member states. In addition to that, factors like the financial crisis, internal political turmoil in the echo of the British exit, or the refugee crisis put additional pressure on the functioning of the union. The EU needs solidarity and cohesion to overcome present and future challenges. Accordingly, it is the dominating goal of the Regional Policy to bring the members economically closer together. The redistributive character of the structural funding is supporting member states and regions in the economic periphery of the EU while contributing to balanced economic growth in the center. Since structural funding is now the major element for the reallocation between donors and recipients it is also a politically sensi-tive object. Shedding light on the effects and the long-run outcomes of the cohesion policy is of particular importance to efficiently manage and control the funding structure.

(4)

The present study can fall back on an assorted set of data to evaluate economic integration and, thus, provides a new concept to investigate the impact of the Regional Policy. That is, the World Input Output Database (WIOD) sets the framework to estimate the fragmentation of in-ternational value chains. By disaggregating the value chains of industries in EU countries, one can show how much value is added domestically and in other countries. More specifically, one can estimate how much value is added in other member countries of the EU. The theoretical considerations in this study suggest that an increase of the share in value added coming from other EU countries and the associated cross-border production linkages are a sign for higher economic integration within the union.

The results of the Input-Output computations confirm previous findings on international

production fragmentation, as forwarded byLos et al.(2015). Within the scope of the dataset

compiled for this study, the average domestic value added shares decrease during the obser-vation period. Moreover, the foreign value added shares are increasingly shifting towards a global production fragmentation. Nevertheless, the cross-border fragmentation of value chains across the EU also accounts for a considerable share of international production fragmenta-tion. Based on these preliminary findings, this study tests the impact of structural funding on the identified value chain fragmentation within the EU. The results of various regression approaches on a cross-section of the data, the entire panel data and different subsamples sug-gest that there is a marginal effect of the cohesion funding, especially in the same or previous year of funding. Withal, the comprehensiveness and validity of the regression results is lim-ited. That is, when considering the structural funding variable with multiple lags, the effect vanishes although the influence of structural funding is expected to have a traceable long-term influence. A tentative interpretation of this discovery is that the problem of reverse causality influences the regression model. Structural funding is often contingent on projects that have to be proposed and issued by entities in the countries. Funds can be seen as framework programs

that can be applied for by local authorities, enterprises or research centers (European

Commis-sion,2006). Therefore, advanced economic integration in terms of higher European production

fragmentation can lead to higher structural funding. An additional test on the impact of spe-cific funds within the structural funding indicates that, even though there are similarities in the effects of the funds, a simultaneous equality of the effects can be ruled out. The sign and magnitude of the impact of the structural funding predictor fluctuates once it is split up into single funds. Despite some limitations, the study offers insights into the connections between the fragmentation of international value chains and the prevalence of structural funding within the framework of the EU Regional Policy.

This study adds twofold to the existing literature. First, contemporary findings on regional

and global value added fragmentation by Los et al. (2015) are extended with respect to an

(5)

Chapter II

Literature Review

This thesis connects two dynamic fields of research. Specifically, the analysis of interna-tional fragmentation of production chains and the evaluation of European cohesion politics. To provide the background for the theoretical considerations and to describe the setting of this thesis, contemporary literature on both research areas is analyzed.

An introduction and overview on production fragmentation is featured inBaldwin(2006),

who reflects that technological advances profoundly influenced the dissemination of produc-tion processes and significantly shaped the new global factory during the second great un-bundling, a term the author uses to describe this new paradigm in globalization. The associ-ated international fragmentation of production processes across borders is an important phe-nomenon in international trade. It has not only a substantial impact on the corporate structure and strategy of multinational enterprises (MNEs), but also on the macroeconomic conditions of the countries that are affected. Thus, it receives a lot of attention in the academic literature.

Early research emphasizing the importance of this phenomenon includes a study byKrugman

(1995), who observes a slicing up of value chains. Similarly, other authors examine the

impli-cations of international production fragmentation and observe a worldwide dispersion of value creation (Antweiler and Trefler,2000;Feenstra,1998;Hummels et al.,1998,2001). One major stylized fact, that surfaced in this early literature on global value chains, is that trade in value added terms provides a more transparent picture of production fragmentation than gross trade figures. While these pioneering papers point out the aspects of this new international trade paradigm and elaborate on the increasingly integrated world trade by carrying out case stud-ies, Koopman et al.(2011) develop a conceptual framework for the modelling of value chain fragmentation. This abstract model enables a complete decomposition of value added and pro-vides a comprehensive description of an economy’s participation in international production. To test the functionality of the model, the authors develop a bilateral database for global trade flows from the OECD Inter-Country Input-Output Tables (ICIO). To avoid the double-counting

in existing international trade statistics,Koopman et al.(2011) identify intermediate goods by

(6)

the database is proprietary, thus access is restricted such that reevaluations of the findings are limited. Second, the tables cover only one year impeding a dynamic analysis. Third, there is no separation between imported intermediates and final goods leading to inaccuracies in the value added composition. The authors emphasize that the creation of a more comprehensive database is indispensable for more sophisticated analyses of supply chain participation.

To-gether with similar studies byDaudin et al.(2011), for an Asian perspective on vertical trade,

and Johnson and Noguera (2012) for value-added ratios in exports, this is the first quantita-tive approach to picture value added decomposition and the international circulation of value chains.

Simultaneously, efforts sponsored by the EU were underway to develop the World

In-put OutIn-put Database (WIOD) by Timmer et al.(2014) to provide a statistical framework that

merges Input-Output tables and international trade statistics with data from national accounts. Ever since the enactment of the WIOD platform in 2012, a growing body of literature is reflect-ing on this rich database to capture the value added distribution across countries and industries to obtain valuable insight into global trade. In the following, a selection of these studies is examined with more detail to obtain a clear picture on the implementation of Input-Output analyses and outcomes of contemporary studies on value added dissemination. This provides the background for the estimation of economic integration and shows the contextual frame-work of this study in the light of the prevailing literature.

Commanding over a multicountry setting with global Input-Output tables from the WIOD,

Los et al.(2015) decompose the value added of specific final goods for all countries involved in the production chain. The focus of the paper is on highlighting recent trends in the inter-national fragmentation of value chains with a focus on manufacturing industries. Their two main findings are that, independent of product type characteristics, there is a higher interna-tional fragmentation of value chains and that these value chains have become more globally dispersed within the observation period from 1995 to 2008. Thus, value added from other than the regional trading bloc has increased to a greater extent. This indicates that, based on the framework and the data of this study, global value chains are more pronounced than

re-gional value chains. This finding is in contrast to the conclusions from studies by Johnson

and Noguera(2012) andBaldwin and Lopez-Gonzalez (2015), who present empirical results opting for an accentuation of regional value chains. Since the present paper focuses on

por-traying the European picture of value added disaggregation, the most important finding ofLos

(7)

regional perspective of the EU. Having new data at hand with the 2016 Release of the WIOD, the data analysis of the present paper for the EU financial cycle 2007 to 2013 aims at shedding new light on these previous findings. In addition, the present paper extends this work by con-sidering the EU cohesion politics as potential determinant of the EU value added shares. Thus, the article implicitly sets the modeling background for the value added fragmentation in this study.

More recently, Pomfret and Sourdin(2016) elaborate on the the differences between

pro-duction fragmentation in Europe and Asia and find that the value chain participation is highly uneven. They highlight that the production unbundling in Europe has been facilitated by the continuous EU enlargement. They argue that the expansions brought in new countries with

lower factor prices that responded to lower trading cost within the union. However, Pomfret

and Sourdin(2016) suggest that only a small number of Central and Eastern European coun-tries managed to become major players in the regional value chain in Europe. The authors argue that participation would require a local environment that fosters the value chain partic-ipation, or rather economic integration into the EU. The present study aims at carving out to what extent the EU can contribute to this environment by initiating structural funding to the member countries.

In addition to the literature on production fragmentation, this study draws on a variety of

studies on the cohesion politics of the European Union. A report by Manzella and Mendez

(2009) provides a good starting point with an overview of the origins and the evolution of the

EU cohesion policy in an historical perspective. Since its enactment, the Regional Policy of the EU went through many qualitative and quantitative changes. The promotion of a harmo-nious development of economic activities throughout the community has been one of the main tasks of the European Commission, and later the EU, ever since the Treaty of Rome in 1957. After the first enlargement at the beginning of the 1970s, the implementation of the Euro-pean Regional Development Fund (ERDF) started the first tangible cohesion policy to correct for regional imbalances and prepared the ground for Economic and Monetary Union (EMU).

Manzella and Mendez(2009) report that, over the last 40 years, the EU cohesion funding has continuously adapted to the economic and societal needs of the union and transformed from a mere budget redistribution instrument to a regional development tool. The reforms prior to the 2007-2013 fiscal perspective align the cohesion funding towards a rapid integration of the new member states that joined the EU in 2004 and 2007. The accession of 12 new members

mostly located in Eastern Europe significantly increased the regional disparities.Manzella and

(8)

With the increasing share of the EU budget dedicated to Regional Policy, a growing body of literature evolved to evaluate the outcomes of the funding policy. Yet, there is ambiguity in the results about the success of the cohesion process. For this reason, the Regional Policy

is still drawing a lot of attention by researchers (Pellegrini et al., 2013). Most inquiries

fo-cus on the implicatoins with economic growth in the receiving countries. While a majority of

studies find a strong positive impact of structural funding (Dall’Erba,2005; Beugelsdijk and

Eijffinger,2005;Maynou et al.,2016;Kyriacou and Roca-Sagales,2012) there is controversy.

That is, the outcomes of other studies vary from a weakly positive impact (Esposti and

Bus-soletti, 2008;Mohl and Hagen,2010) to statistically insignificant findings (Dall’erba and Le Gallo,2008) and even to a negative impact of structural funding (Boldrin and Canova,2001). Likewise, most studies on economic convergence find that the convergece process in the EU is

conditional on various causations like income distribution or labor productivity (Ramajo et al.,

2008). Since these determinants are approached differently in the academic literature, there is

divergence in the outcomes. Therefore, the bottom line on the success or failure of the cohe-sion policy in its numerous fields of activity is ambiguous and to a certain extent remains an open question to research. Despite the vast body of literature, there is no solid consensus in the

factual analysis. A meta-regression analysis by Dall’erba and Fang (2015) demonstrates the

dissent in the prevailing literature but also detects sources of heterogeneity and finds that sev-eral differences in data characteristics are causing ambiguous primary estimates in the studies observed.

Partially, the inconclusiveness of empirical results on the impact of the Regional Policy can be traced back to differences in methodological and data-analytical approaches and vary-ing observation periods. The ex post policy evaluation reports by the EU generally implement micro-, meso- and meta-evaluation studies based on implementation reports or public

consul-tations to measure the success of their policies, most of them thematic in nature (Bachtler and

Wren,2006). In contrast, scholars often build on macroeconomic models on convergence and

growth based on neoclassical theories. Thus, while the empirical approaches by the EU assess the effectiveness of the funding strategy contingent on reaching targets on the regional level, scholars want to test the overall effects of the funding. These differences and the associated incomparability of modeling and empirical structures at least in part account for differences in the outcomes.

Another factor that produces ambiguity results is the nature of the data. It is troublesome to isolate the impact of structural funding in a macroeconomic setting where numerous confound-ing variables influence the outcome variable. On that account, this study aims at explainconfound-ing the impact of the Regional Policy on economic convergence based on the integration of value chains in the EU. Broadly defined, the concept of convergence is the decrease of disparities of

members in a group. AsSachs et al.(1995) argue, economic convergence is connected to the

(9)

countries. Therefore, investigating changes in the European value chain fragmentation sheds light on the integration process of countries in the EU. This opens an entirely new scope for considering the impact of structural funding on economic integration within the EU. Hence, this study uses a different concept to measure the success of the Regional Policy based on a new set of Input-Output data.

So far, there are only a few attempts to quantify European integration with the help of an

Input-Output analysis. One of these studies is brought forward byHoen(2002) who examines

inter-country Input-Output tables of six Western European countries in the period from 1970 to 1985. The major contribution of this study is to demonstrate “how input-output analysis can be used to analyse economic changes of countries involved in a process of economic

integra-tion”Hoen (2002, p.251). The most important result of this analysis suggests that, based on

the inter-country Input-Output tables, there is convergence among the countries in the sample when measured by the sectoral distribution of value added. Disposing over a much more

ac-curate database, the present study aims at overcoming some of the data limitations thatHoen

(2002) addresses in his analysis.

A similar approach by Beutel(2002) uses a dynamic Input-Output table projection for the

fiscal perspective from 2000 to 2006 to quantify the impact of structural funding on economic growth and other indicators. The data is based on an aggregation of harmonized national ac-counts of the Eurostat database and macroeconomic forecasts of the European Commission.

Based on this analysis,Beutel(2002) provides an impact analysis of the cohesion funding by

integrating the structural interventions into the dynamic perspective of the forecasts. The au-thor then compares the Input-Output table to the situation without funding to estimate the net effects in the receiving countries. The quantification of the various structural effects shows large discrepancies in the regional success that defy any generalization about the effect of the funding.

The present study is an effort to connect the papers of Beutel(2002) and Hoen(2002) to

(10)

Chapter III

Hypotheses

The present paper aims at investigating the connection between structural funding and value chain fragmentation. Thus, it is vital to hypothetically connect these two domains. The structural funding as dynamic community policy of the EU is issued and controlled by the Re-gional Policy directorate of the European Commission and subject to budget negotiations for every multi-annual fiscal period. The structure of the different funds that are collectively called structural funding are different by the themes and policy objectives of the particular fund. The Treaty on the Functioning of the EU determines the subordinate goal of the structural funding as the promotion of the "[...]overall harmonious development, [and that] the Union shall de-velop and pursue its actions leading to the strengthening of its economic, social and territorial

cohesion."European Union (2012, Art. 174) This target should be reached by "[...]reducing

disparities between the levels of development of the various regions and the backwardness of the least favored regions."European Union(2012, Art. 174)

To understand the origins and the intention of the Regional Policy, one has to look back into the development of the EU. In a historical perspective, the signing of the Maastricht Treaty in 1992 and the subsequent decision for an internal market paved the way for the EMU. However, large disparities between nations had to be addressed first to establish a coherent monetary union. Therefore, redistributive mechanisms were extended and intensified. This marks the first generation of large-scale community support and structural funding that represents a

gen-uine regional development tool rather than an intergovernmental budgetary transfer (Manzella

(11)

The particular importance of the structural funding was most pronounced in the run up to the common currency area under the framework of the EMU to prepare countries to fulfill the

preconditioned convergence criteria (Manzella and Mendez, 2009).1 By entering the EMU,

countries abandon national monetary policy instruments and, therewith, step back on fiscal policy discretion to intercept asymmetric shocks. Thus, economic convergence as main goal of the Regional Policy, aims at contributing to a balanced EMU by financially supporting the

countries to absorb potential shocks (Maynou et al.,2016).

With the increasing importance of economic convergence within the growing EU, the fi-nances devoted to structural funding gained substantially. In 1992, the budget for structural

operations in social and economic cohesion consisted of 18.6 billion ECUs2and accounted for

about 28 percent of the total budget (Commission of the European Communities, 1992). In

1997, the structural funding already accounted for 33 percent of the budget with 29.3 billion ECUs to allocate. In the late 1990, when the funding decisions for the next budgetary period were set, economic recession, consolidation and uncertainty across the union let the finances for structural funding remain constant at about EUR 30 billion annually for the time from 2000 to 2006.

The prearrangement of the fiscal perspective from 2007 to 2013 which is in the focus of this study can also be described by reforms of the funding structure and adaptation of financial means. That is, after the accession of 12 new countries in the expansion rounds of 2004 and 2007 the preexisting income and wealth gap between the member states increased profoundly. Therefore, the Regional Policy set up a new budget structure and defined new objectives. These implications for the financial structure, the funding allocation and the policy objectives archi-tecture for the focus period are explained in detail in the data analysis section of this paper. In short, the primary aims of the 2007-2013 funding period are the creation of jobs and growth as well as the development of an innovative European economy. The aim is to release the vast economic potential on the regional level, especially in the newly accessed member states in

Eastern Europe, that remained unused so far (European Commission, 2006). The strong

fo-cus on jobs and innovation across the EU requires economic cooperation between the member states. On that account, the European internal market facilitates the free movement of goods, capital and labor throughout the 28 member states. The structural funding is a decisive element in this context. It is the key policy of EU regional integration and conversion measure. Thus, the funds of the Regional Policy are directed at supporting economic cooperation within the common market of the EU. National innovation systems in the all member states, but especially in the poorer regions, can profit greatly from economic cooperation. Likewise, the creation of jobs is also dependent on the functioning of the internal market and European economy.

1_{The insufficiencies of this process, e.g. the inadequate selection and lacking supervision of Greece as an} EMU member, proved to be a major stumbling block for the development of the EU in recent years.

(12)

This is where the connection to the value chain fragmentation is established. The model

that is constructed in the Theory Chapter of this paper proposes a disaggregation of value

added among the member states of the EU. At the bottom line of this model is the differen-tiation between value added that is created domestically, in other EU countries and in other countries of the world. Principally, the fragmentation of international value chains is an indi-cator for the process of economic integration. When less value content is added domestically, the value chain of an industry relies more on the share of value that is added in other countries and trade flows between those countries increase such that cross-border production linkages are generated.

The fundamental proposition of this paper is that the value added shares in value chains of EU member states are sensitive to structural funding on country level. Thus, the aggregated value chain of an exemplary member country, hypothetically, shows increased international production fragmentation due to the acquisition of structural funding. This reflects the con-jecture that structural funding by the EU is directed at economic integration of member states. That is, a decrease in the domestic value added shares of a member state, that can be attributed to the structural funding, implies a positive performance of cohesion funding. The fewer value a country adds to value chain of a domestic industry, the more value enters the value chain from other countries. In the broader picture, this indicates that industries are more integrated which, in turn, leads to the emergence of trade and the creation of jobs, which is at the top of the agenda of the Regional Policy. Generating country group subsamples in the regression analysis, like a focus on newly accessed member states, provides a more differentiated view on the value chain fragmentation between countries. Likewise, sub-sampling in the industry selection can point out differences in the impact of structural funding.

(13)

The third hypothesis of this paper is that the components of the structural funding, which follow subordinated objectives, have a different impact on the value added fragmentation. So far, the structural funds are regarded as a composite number. However, the implementation of the Regional Policy consisting of several funds that provide assistance within the frame-work of cohesion funding. The three major funds considered here are the European Regional and Development Fund (ERDF), the European Social Fund (ESF) and the Cohesion Fund

(CF).3 Therefore, the third hypothesis of this paper is concerned with the differences in the

response to the respective funds. Every fund has preassigned objectives that are penetrated by target-oriented investments. A Council Regulation in the run up to the 2007 to 2013 funding period defines the common principles, rules and standards for the three funds that implement

the cohesion objectives as follows (European Union, 2006). According to the regulation, the

ERDF is aimed at promoting public and private investments to alleviate the regional disparities within the region by supporting programs that prioritizes research, innovation and territorial cooperation throughout the EU. The ESF focuses on the labor market by helping workers and enterprises to adapt and promote a labor market that is equally accessible for workers of all member countries. The CF has a key investment area in trans-European transport networks that promotes the flow of goods between countries. Unlike the ERDF and the ESF, the CF is restricted to economically underdeveloped regions in the union. A more detailed specification

of these objectives is carried out in thedata descriptionfor the empirical approach in this paper.

In this spirit, the third hypothesis suggests that distinct themes in the cohesion funding have varying effects on the measure for economic integration. The aim of testing this hypothesis is to discover whether there are different impacts of the three funds on the EU value added.

(14)

Chapter IV

Theory and Model

1 Measuring Value Added Fragmentation in European Value

Chains

Testing the previously developed hypothetical considerations for this study requires a meth-odological background for the international fragmentation of value chains which is set by ex-isting theoretical and empirical models. More specifically, the preliminary calculations for the data collection on economic integration relies on the identification of the value added per

country in the respective value chain. In the spirit of Los et al. (2015), the model applied to

this setting aims at recognizing the value added contributions of all countries in the sample of the WIOD to the final output of each industry-country pair. Thus, the value of an end product that is produced by a specific industry in a specific country can be decomposed into the value added shares of all other industry-country pairs that contribute to the final value. This measure includes all tiers of suppliers in the production chain. Thus, it decomposes the direct interme-diate input of the first tier as well as all other upstream tiers that contribute to the value of the intermediates that are assembled in a final product.

This measurement scheme is graphically depicted in a simplified representative structure

with a value chain spreading over 4 countries inFigure 1. Imagine a international value chain

(15)

Domestic Intermediates

Final Product

Capital and Labor

Capital and Labor Capital and Labor

Capital and Labor

Country 4

Country 3

Country 2

Country of Completion: Country 1

Intermediates Intermediates Foreign Value Added Domestic Value Added Value Added

Figure 1: Schematic Display of International Value Chain Fragmentation with 4 countries,

Source: Own Display followingLos et al.(2015)

The flowchart inFigure 1considers all upstream economies by disaggregating the value of

(16)

In contrast, using the WIOD enables a more sophisticated disaggregation of the value chain such that value added can also be attributed to countries further upstream in the production pro-cess. More specifically, the aim of the methodological approach in this paper is to point out the country-wise contribution of value added to the final product of each industry i in every country c, thus each industry-country pair (i, c) in the sample. Based on the basic simplicity

as-sumptions of Input-Output calculations as forwarded byLeontief(1953), one industry-country

pair produces only one product. The production process falls back on activities in s = 1, . . . , S industries in n = 1, . . . , N countries. Relating back toFigure 1, the industry-country pair (i, c) is Country 1, the country of completion. The value of the final product relies on domestic intermediates from the country of completion and all other upstream countries k involved in the value chain of that industry, schematically displayed by Countries 2-4. Hence, the final

output value of a good, FO_(i,c), produced by one industry in one country equals the sum of all

value added contributions VA_(i,c)from industry-country pairs in countries k as indicated by the

black bar on the right-hand-side ofFigure 1. This is denoted by the followingEquation (1):

FO_(i,c)=

_∑

VA_(k)(i,c) (1) The aggregated value added for one industry-country pair can be decomposed into foreign and domestic value added where domestic value added is generated in the country of comple-tion and the foreign value added in all upstream countries which is also pictured in the stacked

bar on the right-hand-side of Figure 1. Based on this, Equation (2)depicts the value added

outside of country c, which are all countries but country c, thus (k 6= c). This, in turn, must

equal the final output as computed in Equation (1) minus the value added of that respective

country c.

FVA_(i,c)=

_∑

(k6=c)

VA_(k)(i,c)= FO_(i,c)−VA_(c)(i,c) (2) The share of foreign value added is then generated by dividing the foreign value added by

final output as shown in Equation (3). This step is necessary to point out the importance of

value added in the value chain.

FVAS_(i,c)= FVA(i,c)

FO_(i,c) (3)

Similarly, this model fits to compute the domestic value added shares DVAS_(i,c)of country

kas displayed byEquation (4)andEquation (5).

DVA_(i,c)=

_∑

(k=c)

VA_(k)(i,c)= FO_(i,c)−VA_{(k6=c)(i,k6=c)} (4) DVAS_(i,c)= DVA(i,c)

(17)

In sum, the calculation as displayed by the following equation has to hold:

∑

(k)

VA_(k)(i,c)= DVA_(i,c)+ FVA_(i,c)=

_∑

(k6=c)

VA_(k)(i,c)+

_∑

(k=c)

VA_(k)(i,c)= FO_(i,c) (6) Relating back toFigure 1, the left-hand-side (∑(k)VA(k)(i,c)) and the right-hand-side (FO(i,c))

elements ofEquation (6)show the aggregated value added of the industry-country pair (i, c).

The elements in between show the decomposition between domestic and foreign value added. The next specification of the model aims at identifying the changes in value added shares taken over by other countries in the EU. This needs another disaggregation of the foreign

value added FVAi,cfromEquation (2). That is, the consideration of world Input-Output tables

incorporates other foreign value added from countries outside of the EU in the composition.

On that account,Equation (7)singles out the foreign value added that is taken over by countries

within the EU (EUVA_(i,c)) by excluding the value added contributed by countries outside of

the EU (OVA(i,c)).

EUVA_(i,c)= FVA_(i,c)− OVA_(i,c) (7)

Again, considering the EU value added as share of total final output inEquation (8)points

out the importance of the EU value chain.

EUVAS_(i,c)= EUVA(i,c)

FO_(i,c) (8)

2 Value Added Fragmentation and Cohesion Funding

The second step of approaching the data in this paper consists of a regression analysis to estimate the impact of EU cohesion funding on the value added fragmentation of production chains in EU countries. As developed in the theoretical section of this paper, the basic rea-soning behind this approach is that foreign value added shares and especially EU value added shares are influenced by financial dedications through the cohesion funding. This is because the structural funding is the central instrument of the EU Regional Policy and in pursuit of closer economic cooperation between the member states. On that account, the European struc-tural funding is a catalyst for the EU policy objectives on economic growth and job creation.

To detect the influence of structural funding on the value added shares, a straightforward OLS regression model with the respective value added shares as dependent variable is subject to the problem of non-stationarity. That is, the value added shares and the independent focus variable of structural funding are both trending variables which could result in a spurious

cor-relation.1 Therefore, in order to conceptualize this interdependency between cohesion funding

(18)

and value added fragmentation as a non-stationary progress, following regressions inEquation (9)andEquation (10)are proposed:

DVAS_(i,c,t)− DVAS_{(i,c,t−)}= β+ βSF(c,t−T )+ βvc,t+ λt+ τc+ γi+ εc,i,t (9)

The regression inEquation (9)estimates the impact of structural funding on the changes in

domestic value added shares compared to the preceding period. Thus, the dependent variable is generated by subtracting the prior value added contribution in period t − 1 from the value added contribution in period t. The sum of the dependent variable displays a positive rate when country c increased the domestic value added shares in industry i within the observation time. Vice versa, a negative rate indicates a decrease in the domestic value added shares. The ex-planatory focus variable of the regression, SFc,t−T, consists of the structural funding as share of

current GDP that country c received in year t − T with T = {0, 1, 2, 3, 4} indicating a temporal lag from zero to four years. That is, structural funds are considered as long term investments like infrastructure or social projects that take time to implement and translate into changes in value added shares. Therefore, the lag in the structural funding variable facilitates a dynamic consideration with structural funding data dating back several years. Other explanatory control

variables that enter the regression are pooled in vector vc,t. A detailed specification of these

variables is carried out in the thedata descriptionpart of this paper. Furthermore, country fixed

effects τc, industry fixed effects γiand time fixed effects λt enter the regression to account for

unobserved heterogeneity which transforms the regression into a least square dummy variable

estimator (LSDV). The error term is displayed by εc,i,t. Further diversifications of this model to

consider subsamples of the data are discussed in theRegression Analysissection of the study.

EUVAS_(i,c,t)− EUVAS_{(i,c,t−)}= β+ βSF(c,t−T )+ βvc,t+ λt+ τc+ γi+ εc,i,t (10)

A second regression inEquation (10)offers a different perspective on the value added

frag-mentation. The dependent variable comprises the change in the EU value added shares within the observation period which serves as a proxy for European economic integration. Identical toEquation (9), the dependent variable is generated by subtracting the prior EU value added contributions in period t − 1 from the EU value added contribution in period t in order to

display the changes of EU value added shares. The remainder of Equation (10) is identical

to Equation (9). Thus, this second regression estimates the impact of the (lagged) structural funding variable on the value added share coming from other EU states. It shows the value chain fragmentation among EU countries within the observation period which is fundamental for elaborating the context of this thesis.

In order to specify the impact of the cohesion funding, an additional structure of regression

is proposed in Equations (11)-(13). In every financial perspective, the EU regional funding

is realigned by adjustments in the budget structure to appropriately address the needs of the

respective regions. On that account, a disaggregation of the SFc,t−T variable facilitates a more

(19)

differ-entiations in funding areas can be split up into the ERDF, the ESF and the CF, all of which

comprise specific funding areas that are described in theData Sources. The equations are first

run in a composite regression to test the equality of the coefficients and then in single regres-sions as pictured here to investigate the particular effects.

EUVAS_(i,c,t)− EUVAS_{(i,c,t−)}= β  + βERDF(c,t−T )+ βvc,t+ λt+ τc+ γi+ εc,i,t (11)

EUVAS_(i,c,t)− EUVAS_{(i,c,t−)}= β  + βESF(c,t−T )+ βvc,t+ λt+ τc+ γi+ εc,i,t (12)

EUVAS_(i,c,t)− EUVAS_{(i,c,t−)}= β  + βCF(c,t−T )+ βvc,t+ λt+ τc+ γi+ εc,i,t (13)

Based on this methodological background, the hypotheses that were developed previously are being investigated concerning the effect of the structural funding variable on the proposed measure for international value chain fragmentation. Following tangible hypotheses emerge from the theoretical considerations:

Hypothesis 1: Financial support from the European Regional Policy negatively influences the domestic value added shares of member states estimated during

the financial perspective 2007-2013: The coefficient of the SF_{(c,t−T )} variable in

regression equation(9)is negative and statistically significant.

Against the background of the regression in Equation (9), the first hypothesis suggests

that structural funding is negatively associated with the development of domestic value added shares. This, in turn, suggests that the foreign value added shares increase which is a sign for international value added fragmentation.

Hypothesis 2: Financial support from the European Regional Policy increases the value added shares in other EU countries during the financial perspective

2007-2013: The coefficient of the SF_{(c,t−T )} variable in regression equation(10)is

positive and statistically significant.

Thus, economic integration of member states as estimated by EU value added fragmen-tation shows a positive impulse to incoming structural funding. The decomposed structural

funding, as displayed in Equations (11)-(13), is tested against the background of the third

hypothesis which aims at clarifying whether specific funds under the umbrella of structural funding also have different impacts on the EU value added.

(20)

Chapter V

Empirical Approach

1 Data Sources

At the center of the empirical considerations in this paper is an approach to acquire a clear picture of the value chain fragmentation within the EU based on Input-Output tables. Further, an estimate of the impact of EU cohesion funding on the value added fragmentation within the EU is attempted upon the results of the Input-Output calculations. Thence, the empirical approach relies on two disparate sets of data. First, the decomposition of the EU value chains and the associated assignment of value added shares across member countries draws on data

from the WIOD.1Second, the regression analysis requires data on the structural funding flows

from the European budget (European Commission DG REGIO, 2017). Further, the control

variables in the regression analysis incorporate complementary data on the countries in the sample.

Value Added Shares

Before turning to the Input-Output computations, it is worth describing the WIOD database against the background of the theoretical model to fully understand the scope of the operation.

A simplified world Input-Output table is displayed in Table A1in Appendix I. According to

the groundbreaking work byLeontief(1936), the fundamental rationale behind Input-Output

tables is that the output of an industry relies on factor inputs like capital and labor and in-termediate inputs. Additional factor inputs and inin-termediates are required to produce these intermediate inputs in other countries. The example of the simplified world Input-Output table inTable A1, with two countries A and B and the Rest of the World (RoW), displays this scheme in the upper-left 3x3 matrix with domestic intermediates on the diagonal. Every country has only one industry that requires domestic intermediates as well as domestic factor components like capital and labor plus intermediates from the other country and the RoW . Thus, the prod-ucts of an industry can either be used as intermediates for other industries or be consumed via Final Demand (FD). The final use of the products of an industry are displayed in the second

(21)

3x3 matrix of the world Input-Output table. The end of the columns and the bottom of the rows, respectively, show the aggregated Output of an industry. For this study, the most im-portant part of the exemplary Input-Output table is the Value Added that is generated in each country and added to the final value of a production chain. For the accounting framework to hold, the value added of an industry has to be equivalent to the domestic and foreign final use of that industry’s final product. Thus, this consideration focuses on the value added content dependent on global final demand. In sum, when applying the Leontief model to the account-ing framework of an Input-Output table, inputs on all levels of production, dependent on the global final demand, can be assigned to their sources.

The application of this simplistic framework on the dataset that is used in this study re-quires a preliminary look into the most recent WIOD release from 2016 that covers the period from 2000 to 2014. The WIOD contain information on 43 countries including all 28 EU

coun-tries (as of July 1st, 2013), 15 other major economies and the rest of the world (Timmer et al.,

2016). Within each country and the RoW, 56 industries are classified according to the

Interna-tional Standard Industrial Classification (ISIC). An overview of the industries classified in the

WIOD is given inTable A2in Appendix I.

As to be seen in an exemplary accounting framework of the World Input-Output Tables

(WIOT) in Table A3 in Appendix I, the constellation of 44 countries (when considering the

RoW as a country) and 56 industries leaves 2,464 industry-country pairs. The final output

values as the sum of the columns are the aggregated values of the product of each industry-country pair of completion, respectively. Thus, the values in the bottom row show the sum of all value that is added by all other industry-country pairs that participate in this value chain. Consequently, the sum of these values must equal the World GDP. Based on this setting, the WIOT constitutes an exhaustive outline of the global transactions and trade of intermediates

between countries.2

This framework provides information on the value that is added by an industry-country pair to all global value chains. This value added can be summarized across the rows for each industry-country pair in the second to the last column. To arrive at the value added per country in the last column, the value added of all 56 industries in one country is aggregated and equals the total value added of a country. Similarly, the value added coming from one country can be assigned to the final product of a global value chain. As an example, summarizing the area

highlighted in blue inTable A3, one can obtain the aggregated value added from all industries

in Country 44 to all industries in Country 1. Likewise, the area highlighted in orange sum-marizes the value added from all industries located in Country 1 that is incorporated in the aggregated final products of all industries of Country 44 .

2_{There is one limitation to this calculation: Since the RoW is an aggregation of all other countries but the 43} countries with detailed information, trade between countries in the RoW cannot be accounted for. According to

(22)

Relating back to the diagram inFigure 1, this table provides information on the domestic and foreign value added shares that are incorporated in all industries of a country. That is, domestic value added in Country 1 can be displayed by the area shaded in gray. Then, the foreign value added is calculated by adding the remainder of the column named Country 1 for the Countries 2-44. To obtain the European perspective on value added, the 28 member countries of the EU can be regarded separately while the remaining 15 countries are added to the RoW . For the European Country 1, the EU value added consists of the value that is added from 27 other EU countries. For an industry-specific consideration of value added by other countries, the same exercise can be done for the respective industry column within a country. This displays how much domestic, EU and extra-EU value is incorporated in the final output of a specific industry.

Structural Funding

The second stream of data that is needed for the analytical section of this paper is the struc-tural funding which is determined by the EU Regional Policy. As amplified in the hypothesis section of this paper, the structural funding as EU policy instrument is subject to change ever since it was established. The run-up to the budgetary perspective starting in 2007 is no ex-ception. Most reforms took place in the allocation of resources, the monitoring and control tools and the eligibility of funding areas and projects. Most notably, the cohesion funding for territorial cooperation within the union was increased for the new member states that en-tered the union in 2004 and 2007. Likewise, the amount of funding increased significantly up to EUR 347 billion and 35.7 percent of the total budget which marks the first time that

the structural funding captures the main expenditure commitment in the EU budget (European

Commission,2006). The finances dispense to EUR 201 billion for the ERDF, EUR 76 billion

for the ESF and EUR 70 billion for the CF (European Commission, 2006). The funds are

in-struments for the implementation of the Regional Policy agenda and, therefore, follow a policy architecture aligned towards three main objectives: Convergence, Regional Competitiveness

and Employment and European Territorial Cooperation (European Commission, 2007). The

transition from the 2000-2006 planning period comes with reforms to simplify the funding

structure which is pictured inTable A4in Appendix I. It can be seen from there that the three

funds combine to serve the three main objectives that boil down other objectives and initia-tives from the previous funding period. Being the largest fund, the ERDF is involved as an instrument for every objective. The CF is entirely committed to convergence and the ESF dis-perses over the convergence objective and regional competitiveness and employment objective. The finances for the funding are provisioned budgetary commitments for EU financing, national co-financing and total financing. However, the actual funds paid out to the member states every year are lower. The difference between the funds planned in the budget and the allocated finances is substantial. The newest available data on funding allocation, which is used in this study, shows that about EUR 210 billion of the available EUR 347 billion for the

(23)

Com-mission DG REGIO,2017). This results in an absorption rate of just over 60 percent. A report from the Regional Policy directorate elaborates on difficulties in the financial execution of

structural funding (Bubbico and de Michelis, 2011). The report points out that financial

ab-sorption is a dynamic process rather than an immediate disbursement which explains the slow implementation of the funds. Moreover, the provisioned funds in the budget have to be issued by governments, local authorities or other entities that are eligible for structural funding. To avoid a bias by including funds that are not payed out, the dataset on structural funding that is used in the empirical approach only considers funding that is already allocated. In

accor-dance to that, Figure A1in Appendix II shows a graph with the respective structural funding

payments per country in absolute terms. The funding payments are averaged over the entire funding period and stacked for the three funds. As can be seen from the graph, Poland takes the largest average share of structural funding between 2007 and 2013, followed by Spain and Germany.

To account for the different sizes of the economies in the sample, the average annual

struc-tural funds have to be divided by the current national GDP.3 This leaves average structural

funding shares that are displayed by the diagram inFigure A2 in Appendix II. EU members

with the largest shares of structural funding when adjusted for the economic size of the

coun-try are countries in Eastern Europe and the Baltic. This is also pictured in the map ofFigure

A3. The dark blue areas are the regions where the highest structural funding as share of GDP

occurred. Being the newest members to the union, large shares of the structural funding are

directed towards the East of the EU to facilitate a process of catching up.4 In General, the first

block of countries with funding shares of more than 1 percent of national GDP are all members that entered 2004 or later, except for Portugal.

Control Variables

The vector of control variables, as implemented in Equations (9)-(13), comprises a set of

variables that potentially take influence on the value added shares. Abstaining from including these variables could lead to an omitted variable bias. To capture the state of economic devel-opment of a country, the current GDP is included as explanatory variable. The data is obtained

from the World Development Indicators database of the World Bank(2017) and adjusted for

the respective currency. The reasoning behind this is that the production fragmentation of a value chain is likely to be influenced by the economic size of a country. Larger economies tend to have well established value chain networks to build upon. Thus, controlling for the economic size of a country increases the validity of the results.

The second control variable that is added to the regression equations on country level is the

3_{The data on GDP is drawn from the}_{World Bank}_{(2017) database and adapted for exchange rates with data} from theWIOD(2016).

(24)

expenditure in research and development, which enters the regression with a lag of one year. To test for robustness of the research and development expenditure as control for the innova-tiveness of a country, the national patent applications per 1 million inhabitants and the share of human resource in science and technology are tested and reveal only marginally different results.

Additional unreported tests on the inclusion of industry-specific control variables fall through because of data availability for the entire set of countries. Preliminary tests on the inclusion of Internet and Computer Technology (ICT) investment rates from the EU KLEMS database

provided byJäger(2016) to the regression could not prove to be of importance.5

2 Input-Output Data

As a first step to fill the previous elaborations with data, the value added contributions in each of the industry-country pairs is calculated. Thus, in order to conceptualize the model that

is developed in theory more formally, following equation based onLos et al.(2015) serves as

point of departure:

g= gtier+ gtier+ gtier+ ... = ˆv(I − A)−( ˜Fe) (12)

This fundamental Equation 12computes the value added generated by all industries in all

countries that can be attributed to the value of one industry-country pair. That is, g is the

summation of all tiers gtier0− gtierK _{that contribute to a value chain (i, c). Referring back to}

Figure 1, the first tier is the country of completion, Country 1. The value added of this last

stage is computed by gtier0= ˆv ˜Fe. This equation has three components. First, the value added

vector, ˆv, which constitutes the value added is generated per unit of gross output. The vector

is displayed in a matrix with the vector elements on the main diagonal. Second, the final

de-mand matrix ˜F, where the tilde denotes that this is a selected final demand matrix where only

the elements in the row, that are associated with one specific industry-country pair, contain its values and the rest of the matrix is set to zero. Multiplied with the third component, the

summation vector e, vector ˜Feis generated. It contains one single element for the respective

industry-country pair while all other elements are zero.

The first tier is then calculated by the expression gtier1= ˆvA ˜Fe. That is, the intermediate

products A ˜Fedelivered to the country of completion have to be included in the decomposition.

The value added generated by the second tier gtier2 is then calculated by gtier2= ˆvA(A ˜Fe).

Logically, the value added by all following upstream tiers is then calculated by repeatedly multiplying with the production matrix A. To account for all tiers in the global value chain

more conveniently, the Leontief inverse (I − A)−1is incorporated inEquation 12.

(25)

At this point, the stylized accounting framework in Table A2can be filled with data. As pointed out in the data description, there are 44 countries with 56 industries that add up to

2,464 industry-country pairs. Thus, the entire area of Table A3consists of a matrix with

di-mension 2,464x2,464 that contains the output of all tiers according to global final demand, as

displayed inEquation 12.

The next step is to aggregate the value added of all industries in one country to generate the value that is added by the respective country. Therefore, the obtained matrix with dimen-sion 2,464x2,464 is restructured as a single column vector with 6,071,296 elements . Then, by country-wise addition of all the elements that correspond to one industry-country pair, the value added contribution of each of the 44 countries to each of the 2,464 industry-country pairs is carved out. From there, simply adding all 56 industries of one country provides a 44x1 vector for each of the 44 countries in the sample. The entries in each of the vectors represent the value added by each of the 44 countries to all the final products of the respective country.

From there, splitting up domestic and foreign value added is achieved by singling out the entry that responds to the country in question and summarizing all other entries. For the first country in the sample the first entry of the 44:1 vector responds to the domestic value added, for instance. Dividing this value by the total final output of the country gives the domestic

value added share, as needed in Equation 9. Accordingly, the sum of all other entries in the

44x1 vector determines the foreign value added. As modeled inEquation 7, the specification

of European value added is then done by subtracting the value added by countries outside of the EU from the foreign value added. Dividing by the aggregated value added of a country gives the share of value added of total final output of a country which is needed to determine

the focus variable for the regressions inEquation (10) to (13). This operation is carried out for

all countries in the sample of the WIOD.

Since this study focuses on value chains within the EU, the country sample departs from

this exemplary representation in Table A2. That is, only the 28 EU member countries in the

sample of 44 countries in the WIOD are considered. The remaining 15 economies and the initial RoW constitute to the RoW from a European perspective, thus all countries that are not in the EU. This specification facilitates a disaggregation going beyond the view of domestic and foreign value added and additionally separates the foreign value added into value added by members of the EU and value added by extra-EU countries, the RoW.

In an effort to illustrate the changes of value added composition in specific industries of

specific countries, Table 1 exemplary displays the value added shares of the transport

(26)

with nearly 30 percentage points compared to little more than 10 percentage points for the Ger-man industry. A further decomposition of foreign value added shares into EU and extra-EU value added shares shows that value added for the German transport equipment manufacturing industry is generated within and outside of the EU at approximately equal share in 2007. How-ever, the values for 2013 show a different perspective. While the EU value added shares stay at a constant level, the Extra-EU value added shares increase by more than 10 percentage points. Thus, the transition from domestic to foreign value added in this specific industry-country pair can be accounted to countries outside of the EU.

Similarly, foreign value added shares in the Dutch electrical equipment manufacturing in-dustry were equally distributed between EU and Extra-EU countries in 2007. However, the situation in 2013 is different since value added shares within the EU borders outperform those from outside the EU. This leaves a situation where the majority of foreign value added comes from within the EU in 2013 suggesting strong EU cross-border production linkages in this industry-country pair. The composed dataset in this study includes this exemplary disaggrega-tion for all industry-country pairs in all 28 EU countries with 56 industries each, hence, 1568 industry-country pairs for every year from 2007 to 2013.

Manufacture of other Transport Equipment

Manufacture of Electrical Equipment

Country of Completion Germany Netherlands

Year 2007 ∆ 2013 2007 ∆ 2013

Domestic VA 67.59 -11.25 56.34 69.97 -28.16 41.81

Foreign VA 32.41 11.25 43.66 30.03 28.16 58.19

EU VA 18.44 0.09 18.53 15.42 23.83 39.24

Extra-EU VA 13.97 11.16 25.12 14.61 4.33 18.94

Table 1: Industry-specific changes in value added fragmentation from 2007 to 2013. Own

calculations based onWIOD(2016) followingLos et al.(2015).

Table A5andTable A6in Appendix I provide an overview on the value added shares for this dissemination on country- and industry level, respectively. The shares on the country level are averaged over all industries and the shares on the industry level are averaged over all coun-tries. Both values are weighted with the final output of each product group that is given in the

last column ofTable A6.

Again, taking the Netherlands as an example fromTable A5, the data suggests that, within

(27)

coming from other countries. The same line of reasoning can be applied for interpreting the

value added fragmentation in the industry-specific display inTable A6.

To clarify the extensive illustrations in Table A5and Table A6, the shaded countries and

industries indicate an increase in domestic value added shares during the observation peroid. This shows that a majority of countries and vast majority of industries increased their foreign

value added shares. This relates back to the study by Los et al. (2015), who engaged in a

related approach with a previous version of the WIOD covering the period from 1995 to 2008. They find that value chains are generally becoming more internationally fragmented. The data

sketched inTable A5andTable A6supports this finding on international value chain

fragmen-tation for the selected countries of the EU for the period from 2007 to 2013, however with a slightly lower magnitude. The most pronounced decreases of domestic value added shares occurred in the Netherlands (21.06 points), Poland (16.81 points) and Italy (14.33 %-points). In contrast, the largest increases in domestic value added occurred in Malta (17.03 %-points), Hungary (14.34 %-points) and Luxembourg (13.86 %-points).

A second major finding of the study by Los et al. (2015) is that value added generated

outside of regional blocks like the EU exceeds foreign value added generated within regional

blocks. On that account, Table A7 displays the separation in EU and extra-EU value added

shares for the EU members. More than half of the countries experienced an increase in value added shares coming from other EU member countries from 2007 to 2013, while the other countries saw a decrease in the EU value added shares. The three countries with the highest increase of EU value added shares are also the countries that had the highest increase in foreign value added shares, the Netherlands, Poland and Italy. In the second column, the difference between the Extra-EU value added shares from the year 2007 to 2013 is listed. A majority of countries increased the value added shares coming from the outside of the EU. The magnitude

and dynamics of changes is described by the last three columns ofTable A7. The first two of

which describe the difference between Extra-EU and EU value added shares in 2007 and 2013, respectively. The last column then displays the discrepancy between the differences in 2013 to the differences in 2007. A positive value suggests that the Extra-EU value added shares outperform the EU value added shares within the observation period. Vice versa, a negative value suggests an accentuated increase in EU value added shares compared to Extra-EU value added shares. In total, the evidence is mixed. Slightly more countries experience a global value

added fragmentation. This points to a development that is also observed byLos et al.(2015).

On the basis of a previous version of the WIOD database and for an earlier observation period they not only find that there is an increasing internationalization of value chains but also that the trend is skewed towards global fragmentation rather than regional clusters. Similarly, the

data presented inTable A7 reveals that, in 2013, a larger share of foreign value added shares

are generated outside of the EU as compared to 2007. However, the globalization of

interna-tional value chains is not as evident as in the findings ofLos et al.(2015). There are 10 out of

(28)

the respective countries. The difference to the existing findings can be due to a multitude of reasons, like the financial crisis, that are not being discussed in detail here.

More important for this study is that the updated dataset reveals that, next to the global fragmentation of value chains, there are dynamics in the shares of EU value added. However, at a first glance, the direction of power and the magnitude is inconclusive, as to be seen in the

first column ofTable A7. The scatter plot inFigure A6is an attempt to show the development

of EU value added shares between 2007 and 2013. The plot displays the share of EU value added in the year 2007 and the year 2013. Every dot represents one of 1.505 industry-country

pairs.6 The observations are widespread suggesting that there is a lot of fluctuation in the

val-ues for each industry-country pair. That is, considering the dashed 45-degree line that is shown in the graph, no or little change in the EU cross-border production linkages would result in a clustering of the dots around this line. The inclusion of a regression through the origin aims at showing the approximate trend of the variable. The orange line in the scatter plot accounts for this simple OLS regression by displaying the fitted values of the EU value added shares in 2013 as predicted by the shares of 2007 without a constant. A comparison with the 45-degree line reveals that the fitted values are slightly skewed towards the x-axis suggesting that the EU value added shares in 2013 are higher than in 2007, when seen for all industry-country pairs. This heterogeneity in EU value added shares from 2007 to 2013 suggests that there is a dynamic development. The following regression analysis aims, among other tests, at investi-gating whether and how of structural funding has implications on this fluctuation.

3 Regression Analysis

Before starting with the regression analysis, a preliminary look into the data suggests that there are outliers among the observations. To avoid a bias in the regression analysis these ex-treme values have to be accounted for. As a first measure, all industries with zero value added are dropped from the observations. This concerns two industries that are dropped entirely from

the WIOD sample.7 From a sample size of 1568 industry-country pairs, 1502 full samples and

8 unbalanced samples remain for the regression analysis.

Relating back to the hypotheses setting in the theory section of this study, the first approach is to test whether structural funding has implications on the changes of domestic value added shares of the member states which is specified in Hypothesis 1. To look into this connection, a first regression is run independent of the panel structure of the dataset. The output of

cross-sectional regressions for every year of observation is displayed inTable 2. The results for each

column come from separate regressions with the change of domestic value added share in the

6_{The number of industry-country pairs departs from the full sample with 1568 pairs because it is adjusted for} outliers. For the description of the outlier-detection, please refer to theRegression Analysis.

(29)

respective year, one structural funding variable and the control variables ln(GDP) and R&D. Thus, the presented values in the first column, are the result of 7 separate regressions with the difference between the domestic value added shares from 2013 to 2012 and the respective lag of the structural funding variable as dependent variable. Being the most recent year of obser-vation, the independent variable can be lagged for 6 years, hence the year 2008. The remaining columns report the same regressions for the remaining years with the difference that the lag of the structural funding variable decreases with every year since 2007 is the first year of obser-vation. The inclusion of earlier structural funding shares is not feasible since this would date back to before the financial budget perspective 2007-2013 and, thus, before the reforms.

Table 2: Summation of all Cross Sectional Regressions - Structural Funding and Domestic Value Added Shares

∆DVAS ∆DVAS ∆DVAS ∆DVAS ∆DVAS ∆DVAS

2013 2012 2011 2010 2009 2008 SF 0.103 0.138∗∗ 0.100 0.111 -0.283∗∗∗ 2.915 (0.085) (0.061) (0.094) (0.129) (0.085) (2.014) SFt−1 0.078 0.131 0.088 0.053 0.156 -0.637 (0.091) (0.086) (0.088) (0.109) (0.372) (3.042) SFt−2 0.121 0.202∗∗∗ 0.133∗ 0.315 0.621 (0.098) (0.069) (0.072) (0.438) (0.603) SFt−3 0.096 0.175∗∗∗ 0.133 0.171 (0.093) (0.061) (0.312) (0.710) SFt−4 0.045 0.799∗∗∗ 0.086 (0.075) (0.291) (0.481) SF_t−5 0.289 1.037∗∗ (0.369) (0.474) SFt−6 0.125 (0.632) Observations 1505 1505 1505 1505 1505 1505

Standard errors in parentheses; White-Huber heteroskedasticity robust standard errors. Notes:Regressions are run separately with every lag of the independent variable. Control variables are not reported.

∗_p_{< 0.10,}∗∗_p_{< 0.05,}∗∗∗_p_{< 0.01}

The results suggest that only the difference in domestic value added shares from 2011 to 2012 show a statistically significant impact of the structural funding variable from the same year and with a lag of more than one year. The sign of the coefficient is somewhat surprising. Based on the underlying cross-sectional regression, structural funding payments increase do-mestic value added shares. The magnitude of the impact is increasing with temporal distance to the year of the payment. The data suggests that a minimum increase of 0.138 percentage points in the difference of domestic value added share by an increase of structural funding of one per-cent in the same year and a maximum increase of 1.037 perper-centage points in the variable with 5 lags. A look into the data reveals that the magnitude of the impact of structural funding on the difference in domestic value added shares is marginal at best. That is, the maximum value of

SF lies at 3.978 percent of national GDP per year and the mean is across all observations is at

European Economic Integration and EU Cohesion Policy